sherpa-onnx Releases

Upgrades `@flowglad/node` to 0.29.0 and migrates internal naming from `catalog` to `pricingModel` for better type safety, while maintaining deprecated aliases for backward compatibility.

Jan 18, 2026

This release adds Dart API support for FunASR Nano, improves JavaScript async APIs for offline recognition, and includes documentation improvements for Node addon.

Jan 15, 2026

Added support for Nemotron Speech Streaming and fixed build issues for various platforms.

Jan 14, 2026

Added Google MedASR and FunASR-Nano support, and added Paraformer ASR support for Qualcomm NPU.

Jan 12, 2026

This release fixes a bug where the app router automatically adds the `transfer-encoding: chunked` header to backend requests. It also includes internal fixes for build types and logging errors, along with the addition of e2e tests and documentation symlinks.

Jan 11, 2026

Added Ascend 910B4 model export, ZipVoice TTS support, and Fun-ASR-Nano-2512 support.

Dec 17, 2025

This release provides binary files for QNN (Qualcomm Neural Processing Unit) SDK 2.40.0.251030, specifically targeting SenseVoice ASR models for Android devices like the Xiaomi 17 Pro. It includes necessary .so libraries and documentation for deployment.

Dec 9, 2025

Version 1.12.19 focuses on stability and cross-platform support, fixing bugs in ZipVoice tokenization, NeMo speaker embeddings, and Matcha TTS. It adds support for Spacemit RISC-V, AXERA NPUs, and token-level confidence scores for transducers.

Dec 5, 2025

Version 1.12.18 expands NPU support significantly, adding Qualcomm QNN and Huawei Ascend NPU support for various models like SenseVoice, Zipformer CTC, and Paraformer. It also exports the OmniLingual ASR CTC 1B model and fixes a crash when reading non-wav files.

Nov 27, 2025

This is a minor release tag for v1.12.16, serving as a changelog container without specific new features or fixes listed in the body.

Nov 13, 2025

Version 1.12.16 is a major release adding comprehensive Omnilingual ASR support across multiple languages and frameworks, alongside significant NPU (Ascend, QNN) and TTS improvements.

Nov 13, 2025

This release provides documentation and links for using Sherpa-ONNX models on Huawei Ascend NPUs.

Nov 10, 2025

This release details the constraints of QNN (Qualcomm NPU) regarding dynamic input shapes, explaining that models are limited to fixed durations (e.g., 10 seconds) and are padded or truncated accordingly.

Nov 10, 2025

Version 1.12.15 introduces audio tagging APIs, expands TTS models (Piper, Parakeet TDT), fixes build issues for various platforms, and adds RKNN and Ascend NPU support for Paraformer models.

Oct 22, 2025

Version 1.12.14 focuses on stability and new language bindings, adding Dart API for Spoken Language Identification and providing pre-compiled wheels for CUDA 12.x. It also fixes TDT decoding and adds a simulated streaming ASR example.

Sep 18, 2025

Version 1.12.13 adds RKNN (Rockchip NPU) support for SenseVoice non-streaming ASR models and fixes an issue with initializing the symbol table for the OnlineRecognizer.

Sep 12, 2025

This release focuses on fixing build issues for RISC-V and Android ARMv8l architectures, alongside improvements to Cantonese VITS text-to-speech models. It introduces new support for streaming Russian ASR models via the T-one project and expands language bindings for various non-streaming models.

Sep 10, 2025

Adds new TTS models including Piper and Zipvoice, fixes build/release issues, and adds support for BPE models with byte fallback.

Sep 1, 2025

Adds streaming ASR models for Russian and German, exports Whisper models, and introduces punctuation and TDT duration APIs.

Aug 25, 2025

Expands TTS capabilities with more Piper models, adds Swift API for speaker embeddings, and exports Parakeet TDT models.

Aug 16, 2025